Use of Updated Census First-Name and Surname Data to Improve Race and Ethnicity Imputation - Summary - MDSpire
Coming Soon: Introducing MDSpire News. Learn more
Conexiant’s news site is now MDSpire News. Learn more

Improving Race and Ethnicity Imputation With Updated Census First- and Last-Name Data

  • By

  • Steven C. Martino

  • Anthony J. Sisti

  • Zach Dietrich

  • Ann Haas

  • R. Mitchell Thomas

  • Mark W. Friedberg

  • Marc N. Elliott

  • November 20, 2026

Share

Objective:

To compare the concordance of updated mBIFSG 2.0 and earlier mBIFSG 1.0 estimates with self-reported race and ethnicity.

Approach:
  • Study population: Cross-sectional analysis of 771,215 BCBSMA commercial or Medicare members surveyed in 2025 who self-reported as Asian American and Pacific Islander, Black, Hispanic, or White.
  • Comparison: mBIFSG 2.0 used 2020 Census first-name and surname data; mBIFSG 1.0 used earlier first-name and 2010 Census surname data. Both versions used the same 2020 Census block-group data.
  • Evaluation: Concordance with self-reported race and ethnicity was assessed overall and by sex using C-statistics.
Key Findings:
  • mBIFSG 2.0 concordance increased for all four groups: Black, 0.930 to 0.948; Hispanic, 0.916 to 0.930; Asian American and Pacific Islander, 0.967 to 0.976; and White, 0.912 to 0.921.
  • Improvements were generally greater among female than male members, ranging from 1.1 to 1.8 points for female members and 0.6 to 1.6 points for male members.
  • C-statistics remained somewhat higher for male than female members among Asian American and Pacific Islander, Hispanic, and White groups.
Interpretation:

The authors report that incorporating 2020 Census first-name and surname data improved concordance with self-reported race and ethnicity. They note that mBIFSG probabilities are intended for population-level assessment, not individual classification, and that downstream analyses should account for imputation uncertainty.

Limitations:
  • The study included respondents insured by one insurer, whose membership was predominantly, but not exclusively, located in Massachusetts.
  • Performance may differ among insurers and populations with different racial, ethnic, or geographic characteristics.
Conclusion:

Updated Census name data improved mBIFSG concordance for the four assessed groups in this insured survey population.

Sources:

Original Source(s)

Related Content